[net.lang.c] More powerful tools than CPP, or, Why Change What We've Got?

gnu@sun.uucp (John Gilmore) (10/04/84)

At one point I needed something like the ragged-array example that
was proposed using new CPP extensions.  Being new to the Unix
game, rather than extending CPP I wrote a small program which
#include-d the original initializer (which was a font description),
mashed the data in various ways, then printf'd a replacement
#include file for the "real" program to use.  Voila!  All the power
of an enhanced CPP, brought to you by a real live C program!

We have the ultimate tool under our fingers (a realo trulo programming
language), let's just standardize what we are already using (no wierd
ways of catenating strings, no changing the rules for quoted #define
expansions, etc -- NO CONVERSION!) and get on with it.

Dave Patterson likes to say about his RISC machines:  "If adding this
feature slows the machine by 10% it had better make the machine use 10%
fewer instructions" (paraphrased).  I estimate the benefit of "fixing"
CPP as low and the cost of converting all the programs that used
"outlawed" or "tasteless" CPP features as high.